Release notes

The published TradingView version before this one was v3.4.1, so this release carries everything since: the v3.6.0 rebuild described in the v3.6.0 note, two fixes, and a new set of defaults. The full settings table, both equity curves and the holdout figures are on the v3.6.1 version page.

New defaults

Sixty defaults moved, in five groups (the full list):

• Targets and stops in % of price. Asia long 0.175% / 0.33%, short 0.225% / 0.33%; London long 0.171% / 0.322%, short 0.213% / 0.33%. A fixed number of points drifts as the index moves; a percentage does not.
• Breakeven and trail re-tuned per leg, in %. Each of the four legs was tuned on its own at 0.01% steps. The Asia long breakeven moves the stop to just below entry — it cuts the loss once a trade has shown it can move, instead of turning small pullbacks into scratches. London’s trails arm early and sit wide.
• Max Extension on — Asia 3.275 standard deviations, London 2.15. It skips entries when price is already stretched far from VWAP.
• Exit if IN PROFIT on — Asia at 0.20% ahead after 12 bars, London any profit after 8 bars.
• Asia MACD slow length 14 → 18.

Two fixes

• An entry decided on the last bar of a trading window was labelled CLOSED when it filled on the next bar, which put it on New York’s bracket and end-of-day clock and flattened it a bar later. It now belongs to the session that decided it.
• The stop-exit label now says what moved the stop: Exit Breakeven when only the breakeven did, Exit Trail Stop only when the trail did.

Measured

Micro Nasdaq, 5-minute, 1 contract, $50,000 account, $0.62 per contract per side and 1 tick of slippage. From a TradingView trade-list export taken 25 September 2026, the last 103 days being the window the defaults were chosen on:

window	trades	net	win	profit factor	drawdown
Last 365 days	617	+$14,226	72.8%	1.81	$1,824
Last 103 days	200	+$10,914	84.0%	4.86	$317

Against the v3.4.1 defaults over 1 January 2024 to 22 September 2026, on the Python engine: +$14,896 against +$9,418, profit factor 1.40 against 1.18, maximum drawdown $1,666 against $4,366, and eight of eleven quarters positive against five. On the windows before the tuning period it earned +$1,639 in the year to September 2025 (v3.4.1: −$917) and lost $557 in January – September 2024 (v3.4.1: −$1,401) — better, but not strong. Size by those holdouts, not by the last 103 days.

The trade-off: win rate falls about five points. The old breakevens turned reversing trades into scratches that counted as wins; the new ones let more trades reach target and cut the losers earlier.

If you are on v3.4.1

TradingView keeps the values you saved, so the new defaults do not reach your chart until you press Defaults in the settings dialog or add the script fresh. And because v3.6.0 moved each unit out of its label into a dropdown, your saved target, stop, breakeven and trail values reset when you update — TradingView matches saved values by label. Save Parameters or write them down first if you have tuned them. Details in updating from v3.4.1.

Defaults only: every input and every line of trading logic is the same as v3.6.1. Ten London defaults moved. The full settings, both equity curves and the holdouts are on the v3.6.2 version page.

London ends earlier

London now takes entries from 3:00 to 5:05 AM ET and is flat by 7:00 AM (was 3:00 – 10:00, flat 10:35). Nothing is held past the 7:05 bar, so the New York pre-open and open are free for anything else you run. Both times came from a 5-minute grid (details): the last-entry time is what matters, peaking around 4:45 – 5:10; the flatten time barely matters once it is after 6:30.

London re-tuned for the shorter window

• Short target / stop 0.213% / 0.33% → 0.28% / 0.28%
• Short breakeven at 0.18% → +0.09% becomes at 0.16% → +0.08%
• Short trail 0.27% once 0.05% ahead becomes 0.22% once 0.03% ahead
• Min VWAP slope 14 → 12 points over 16 bars

Measured

Micro Nasdaq, 5-minute, 1 contract, $50,000 account, $0.62 per contract per side and 1 tick of slippage, from a TradingView trade-list export taken 25 September 2026:

window	trades	net	win	profit factor	drawdown
Last 365 days	430	+$10,658	71.4%	1.92	$1,199
Last 103 days	142	+$8,480	85.2%	6.47	$236

Against the v3.6.1 defaults over 1 January 2024 to 22 September 2026, on the Python engine: +$11,254 against +$14,896, profit factor 1.50 against 1.40, maximum drawdown $1,071 against $1,666, worst quarter −$351 against −$690. It makes about a quarter less: much of v3.6.1’s late-morning London profit came from trades running into the 9:30 AM New York open, which this version gives up. Size by the holdouts, not by the last 103 days.

If you are on v3.6.1

TradingView keeps the values you saved, so the new defaults do not reach your chart until you press Defaults or add the script fresh. No label changed, so nothing resets on its own.

US100 (CFD) configuration

The v3.6.2 page now carries a second tab for the US Tech 100 CFD: the same sessions and exits, with four entry filters re-tuned because a CFD feed’s volume is spread across the day where the futures’ is concentrated in New York. On TICKMILL:USTEC at 2 units (the exposure of 1 MNQ): last 365 days +$11,688 at profit factor 1.88, last 103 days +$8,180 at 4.19 — against +$10,471 / 1.75 and +$6,829 / 2.95 for the untuned settings on the same chart. It was tuned on Tickmill; chart it on the feed that matches your broker’s prices, and tune it to that feed for the best results.

Tokyo Drift now ships a spot gold configuration beside the two Micro Gold ones. It is not a new script — load it onto Tokyo Drift through Load parameters, copied from the XAUUSD tab on the v2.2.5 page.

It is also not the Micro Gold configuration pointed at a different symbol. Spot gold and Micro Gold futures share an underlying but not a tick, a volume profile or a session structure, and because VWAP is volume-weighted, the same prices with a different volume series produce a different indicator. This one was tuned against a real XAUUSD chart from the start.

The finding: the configuration was strangling its trades

Fifteen exhaustive parameter grids, about thirty thousand combinations, run against a v2.2.5 gold configuration that was already live on a chart. The result was not a scatter of unrelated improvements — it was one thing said eleven ways. Every protective mechanism except the chop filter was closing trades that went on to pay.

Three come off entirely: the MAE exit, the long trail, and the long time gate. The rest give what remains more room — the long bracket widens, the short target tightens while its stop widens, the short trail triggers later and follows looser.

The largest single change is the VWAP slope gate, from 1.45 to 2.25. It stops the strategy trading while VWAP is essentially flat, which is the condition it has no edge in. That one setting is worth more than every exit change combined, and every value from 2.00 to 2.50 beats 1.45 on both holdout windows — a plateau, not a lucky number.

What it does

Against the configuration it was derived from, same chart, same 365 days:

	parent	XAUUSD
net	+$19,203	+$24,105	+25.5%
maximum drawdown	$2,729	$1,688	−38%
profit factor	1.492	1.622
win rate	62.04%	67.56%
trades	598	447	−25%
net ÷ drawdown	7.0×	14.3×

Spot gold on the OANDA CFD feed, 5-minute, quantity 10, a $50,000 account, the real broker commission and one tick of slippage. Spread is not modelled — at this size it would cost roughly $3.60 a round turn against the $0.50 of commission that is.

What was rejected

A list of wins without the losses is not evidence.

The MACD settings and both RSI blocks were swept exhaustively and left exactly as they were. A faster MACD beat the original on both holdout windows while every other setting sat at its starting value, and then made things worse once the exits were unconstrained — which is the whole reason a combined result has to be measured rather than assumed. A looser overbought level produced eleven positive quarters out of eleven and a deep-history profit factor of 1.78, on 363 trades against 788; a perfect record on a halved sample is the least durable result in a study, not the most.

How far to trust it

The 365-day and selection-window figures are a real TradingView export, and the Python engine twin reproduces it at 122 of 122 entries — 100% recall and precision — with a $25 difference in net profit across $9,397.

The holdout figures are weaker evidence and are labelled as such on the version page. TradingView will not serve more than about 101 days of the OANDA 5-minute series, so there is no way to test an out-of-sample window on the chart’s own data; those numbers come from a Dukascopy spot-gold series that shares only 68% of its entry bars. It ranks configurations reliably. It does not quote what the chart would have made.

Prop Safe is the configuration for funded evaluations, prop accounts and small accounts, where the drawdown limit matters more than the last dollar. The v2.2.4 Prop Safe was built on the previous defaults; this one is built on the current v2.2.5 defaults, with the same exits, the same session and the same time gate. It is not a new script: load it onto Tokyo Drift through Load parameters, copied from the Prop Safe tab on the v2.2.5 page.

Three settings change

• Anti-liquidation filter 3.5 → 2.0 × ATR. Skips an entry straight after an oversized candle.
• Maximum VWAP extension 2.7 → 2.2 standard deviations. Skips an entry when price is already stretched far from VWAP.
• Short RSI reset 39 → 56. A short now needs a real bounce first rather than any dip below the midline.

Everything else is the Default’s.

What it trades off

About half the trades go. The ones that remain win more often and lose less.

Micro Gold, 5-minute, 1 contract	Default	Prop Safe
Last 365 days, TradingView	523 trades, 62.9% win, PF 1.64, +$20,558	270 trades, 68.2% win, PF 1.96, +$12,658
7 June 2026 onward, TradingView	159 trades, 75.5% win, PF 3.21	80 trades, 81.3% win, PF 4.58
January 2024 onward, maximum drawdown	$1,714	$1,140
Quarters positive since January 2024	8 of 11	10 of 11
Worst quarter since January 2024	−$850	−$310

It is not the Default with shorts switched off — it still takes about one short for every three longs, and a shorts-off version measured worse on every window.

How it was chosen

The three settings were the only ones, out of about 2,900 combinations, that raised the win rate without lowering profit factor or raising drawdown on the recent window and on both older windows, 2025 and 2024, at the same time. Neighbouring values hold up as well, so none of them is a knife-edge.

Because those older windows were used to choose it, they are no longer an untouched test for this configuration; the forward test starts now. These are backtests on one instrument, and past results do not predict future ones.

The published TradingView version before this one was v2.2.4. Nothing about the strategy’s structure changed: no input was renamed or removed, so every saved v2.2.x configuration still loads. What changed is two new settings and nine defaults, taken from the configuration actually being traded. The script is Tokyo Drift on TradingView, and the full settings table is on the v2.2.5 version page.

The new setting

A directional time gate stops one side taking new entries for the late part of a session while the other side trades normally. An open position is never touched — it still runs to its own target, stop or end-of-day close. The window runs to the end of that session’s own trading window, so on an overnight session it cannot spill into the next evening’s open.

Long entries are blocked from 12:45 PM ET by default. Short entries are not.

The asymmetry is the interesting part. Gold tends to slump or sell off into the New York close, so the obvious move is to stop buying into it — and that is what the measurement supports. What the measurement does not support is the mirror image. Late shorts are the weaker side on a per-trade basis, but removing them made the account worse: less profit and more drawdown, because those trades were offsetting losses elsewhere in the book. A trade can be poor on its own and still be worth holding for what it does to the shape of the equity curve.

The cut-off is not a fitted number either. Every setting from 10:30 AM to 2:45 PM produces a profit factor of 1.41–1.42 across the full test period, which is the reason to believe it — a single standout time of day would have been a fit.

Breakeven comes off

The larger change is that breakeven is now off on both sides, with the trail set slightly wider and doing the protecting on its own. The long target comes in from 0.52% to 0.375% so winners are banked sooner, and the minimum-VWAP-distance filter is on.

This is a genuine trade-off and worth stating plainly rather than burying. Win rate falls from 61.7% to 58.7%. Breakeven is what used to convert a reversing trade into a scratch; without it those become small losses. In exchange, profit factor goes 1.34 to 1.42, maximum drawdown falls 23% from $2,226 to $1,714, and the number of positive quarters since January 2024 goes from six of eleven to eight of eleven, with the worst quarter improving from −$1,084 to −$850.

If win rate is the number you select on, that is a real cost and you may prefer the previous configuration. If drawdown is what caps your position size, the smaller figure is worth more than the three points.

One visible consequence either way: the Flatten Breakeven exit label no longer appears in the trade list, because breakeven no longer fires.

The short entry finally filters

The short side’s RSI reset threshold moved from 34.5 to 39. At 34.5 that condition was satisfied on 95.7% of session bars — it was not filtering anything, and the short side was entering on little more than RSI below 50. This is the kind of defect that hides well: nothing errors, the strategy trades, and the setting looks deliberate in the panel.

Prop Safe and Split Sessions

Update, 23 September: Prop Safe now has a v2.2.5 version, built on these defaults — see the Prop Safe note and the v2.2.5 page.

Split Sessions stays on its v2.2.4 settings and remains on the v2.2.4 page, which is now marked superseded. Its Load Parameters file was saved before the time gate existed, so loading it onto the updated script leaves the gate at whatever the chart already has. There is a note on that page explaining what to switch off.

Measured

Micro Gold, 5-minute, 1 contract, a $50,000 account, $1.20 per side and 1 tick of slippage. Last 365 days on TradingView: 523 trades, +$20,558, 62.9% win, profit factor 1.64, drawdown $1,665. On the full 1 January 2024 to 21 September 2026 span the engine twin — verified against that export at 95.4% entry-by-entry — measures 904 trades, +$20,765, profit factor 1.42, drawdown $1,714.

These are backtests on one instrument over one period, and the settings were chosen with knowledge of that period. Size by the holdout figures on the version page, not by the selection window.

v3.6.0 was never published on its own. It ships on TradingView as part of v3.6.1, which does change the defaults — “no default moved” below is true of v3.6.0, not of the version you will install. Everything else in this note still describes what v3.6.1 contains.

The published version before this one was v3.4.1. This release carries everything since, and no default moved: an untouched chart trades exactly as v3.4.1 did. Settings you have changed are a different matter — read If you are on v3.4.1 before updating.

One layout, shared with Tokyo Drift

Drift VWAP Pullback tunes each session on its own; Tokyo Drift tunes once for the whole day. They are siblings, and now they read like it. Inside each of the three session panels the groups appear in the same order, with the same names and the same wording, as Tokyo Drift’s global groups:

VWAP · 1-Hour Momentum · Efficiency Ratio · MACD Histogram Velocity · RSI Overbought/Oversold · Max Extension · TP / SL · Breakeven / Trail · VWAP Re-cross Exit · MAE / MFE Stop · Reversal Engine

If you know where a setting lives in one strategy, you know where it lives in the other. Every tooltip was rewritten to say what the setting does and when you would change it, and nothing else.

Points, ATR or percent — everywhere

TP/SL Mode in each session chooses Points, ATR Multiple, % Price Change or Risk:Reward Multiple. In Risk:Reward mode you size the stop (in any of the three units) and the target is that distance times an R multiple, set per side. In ATR mode a Min Points Floor and Max Points Cap keep a very quiet or very violent ATR reading from producing a bracket you would never have chosen.

The Reversal Engine has its own mode, and its breakeven and trail follow it, so a percent reversal bracket no longer rides a points trail.

Every breakeven and trail value — BE Offset, Trail Trigger, Trail Distance, long and short — has its own unit beside it. The Min VWAP Slope and the anti-liquidation Max Candle Range can each be given in points or as a multiple of ATR, which keeps a filter at the same strictness as volatility changes.

Percent of price is the one to reach for if you run the same settings across months in which the index has moved a long way: 85 points was 0.35% of the Nasdaq a year ago and is 0.29% now.

MAE / MFE Stop, per session

Two independent exits in every session. Exit if IN LOSS closes a trade that has been open at least N bars (or minutes) and is still at least X underwater, at the next open, ahead of the full stop. Exit if IN PROFIT closes a trade that has been open at least N bars and is at least X ahead but has stopped making progress. Distance in points, percent of price or a multiple of ATR; time in bars or minutes.

Both read the current unrealised move at each bar close, not the worst or best excursion since entry. A distance of 0 means any amount; a time of 0 switches that side off. The ATR is this group’s own, on its own period, frozen at the fill. Neither side ever starts a reversal. Both ship off.

VWAP Re-cross Exit, per session

Closes a trade once price has closed back on the wrong side of the session’s VWAP for a set number of bars in a row. It uses the VWAP of the session that opened the trade, so a position that outlives its window is still judged against its own anchor. Ships off.

Reversal Engine

Two new switches per session. Require a confirming candle is what the engine has always done — the reversal only enters if the bar at the end of the wait closes in its direction — and can now be turned off. Close a reversal when its session’s window ends flattens a reversal at the end of its window instead of holding it to the end-of-day close.

Global

• Enable Longs / Enable Shorts (global master) — turn a side off everywhere, reversals included, without touching three panels.
• Session display — a box that tracks each session’s high and low, the old full-height shading, or nothing. Each session has its own colour.
• Colour bars during a reversal trade.

Webhooks

• Contracts (qty field) accepts fractions (0.01 and up), for CFD and spot venues that size in lots.
• The sl and tp sent with every entry are worked out in the session’s real unit, whichever TP/SL Mode is selected.
• The strategy name sent with each webhook carries the running version automatically. Leave Strategy Name override blank.

If you are on v3.4.1

TradingView keeps a saved setting against its label. Labels that carried a unit — Long TP (Points), Trail Distance (pts), Max Candle Range (Pts) and their relatives — have lost it, because the unit is now a dropdown beside the value. On update, those settings fall back to the defaults:

• Long / Short TP and SL
• Long / Short BE Offset, Trail Trigger, Trail Distance
• Min VWAP Slope Delta, VWAP Slope Lookback Bars, Max Candle Range, ATR Period
• every Reversal Engine TP, SL, Enable BE, Enable Trail and direction toggle

If you run the defaults, do nothing. If you have tuned any of these, write your values down (or use Save Parameters) before updating and re-enter them afterwards. Everything else — windows, filters, breakeven triggers, guardrails — keeps its saved value.

The published TradingView version before this one was v1.5.1. This release carries everything from v1.5.2 through v2.2.4, and the defaults have moved, so read the If you are on v1.5.1 section before anything else. The script is Tokyo Drift on TradingView; the three configurations and their settings files are on the version page.

The MAE / MFE stop

A new exit group with two independent sides. Exit if IN LOSS cuts a trade that has been open at least N bars (or minutes) and is still at least X underwater, at the next open, ahead of the full stop. Exit if IN PROFIT takes a trade that has been open at least N bars and is at least X ahead but has stopped making progress. Distance in points, percent of price or a multiple of ATR; time in bars or minutes. The ATR is this block’s own, on its own period, frozen at the fill, so retuning the bracket’s ATR never re-scales these exits.

Both sides test the current unrealised move at each bar close, not the worst or best excursion since entry. A zero time switches a side off. Neither side ever starts a reversal.

The default runs both sides: 0.35% of price after 22 bars in profit, 0.35% after 16 bars in loss. On the year-long export they took 68 and 44 of 573 exits, and with the breakeven they are why the largest loss on the year is $217 on a 0.45% stop.

New since v1.5.1

• Percent of price for the trail and the breakeven offset. The bracket has been percent-denominated since v1.3.0; the trail and offset knew only points and ATR, so a configuration could size its stop off price and its trail off ATR with no way to make them agree.
• P&L Time Stop. One bar limit for trades in profit, another for trades in loss. Off by default.
• Held Cross entry mode. Fires on the bar where the close has stayed on one side of the session VWAP for exactly N bars since crossing there. Off by default.
• VWAP Re-cross Exit. N consecutive closes back through the session VWAP flatten the trade at the next open. Off by default.
• Percent of price for the reversal bracket, with the reversal’s breakeven and trail following it.
• Ghost name on the dashboard, reading the exact strategy name the webhook payload carries.
• Session display. Each session panel has its own colour; choose the background shading (default), a session box tracking the window’s high and low, or off.
• The dashboard stays on screen after a settings change instead of going blank until the forming bar closes.

Defaults have moved

The default is the operator’s launch configuration, exported from the chart and reproduced by the engine at 96.2% entry-by-entry. Against v1.5.1:

• One all-day session. Asia is the only session enabled and runs 6:45 PM to 3:00 PM ET; London and New York are off. The VWAP is anchored to the Globex open, so one continuous VWAP carries the whole trading day instead of re-anchoring at each session.
• Brackets are percent of price: long 0.52% target against a 0.45% stop, short 0.57% against 0.45%. v1.5.1 ran 3.5x / 2.35x ATR and 3.0x / 2.95x.
• Breakeven and trail on, in percent of price; both were off in v1.5.1. Reversal engine off; v1.5.1 shipped it on. MAE / MFE stop on, both sides.
• Filters: Efficiency Ratio on at 0.03 with the fade veto, Anti-Liquidation on at 3.5x ATR, Max Extension on at 2.7 standard deviations, RSI filter on as before. MACD Histogram Velocity off; it was on in v1.5.1, and the Prop Safe configuration turns it back on.
• Entry: RSI Reset, 54 long / 34.5 short, 3-bar lookback, minimum slope 1.15 points over 16 bars.

Two more configurations, Prop Safe and Split Sessions, are on the version page as Load Parameters files with their own year-long results. Prop Safe is the default plus the MACD filter, fewer trades and the least drawdown on the long history, and is the one to run on a funded evaluation or a small account. Split Sessions is the three-session shape with the highest net on the year.

Readable orders and a rewritten dialog

Every order on the chart and in the trade list now says what it is: Enter Long, Enter Short, Flatten Target, Flatten Trail, Flatten Breakeven, Flatten Stop, Flatten EOD, MAE Exit, MFE Exit and so on. The webhook payloads are unchanged.

Every tooltip has been rewritten for the trader with the dialog open. The MAE / MFE unit dropdowns carry real titles, and the Distance Mode and trail unit dropdowns are no longer greyed out while their switch is off, so a unit set before the engine is enabled is not lost.

If you are on v1.5.1

TradingView does not reset your inputs when a script updates. Use Reset settings in the strategy’s settings dialog, or remove the script and re-add it; otherwise you are trading v1.5.1’s three sessions and ATR brackets with the reversal engine on, under a script that expects none of them.

Measured

Micro Gold, 5-minute, one contract, $50,000 account, $1.20 per side and one tick of slippage. Each configuration is a TradingView trade-list export over the last 365 days, 16 September 2025 to 15 September 2026:

configuration	trades	net	win rate	PF	max drawdown
Default	574	+$18,542	65.7%	1.52	$2,044
Prop Safe	457	+$15,665	65.9%	1.57	$1,821
Split Sessions	695	+$21,279	52.9%	1.40	$2,643

Outside that year the engine measures the same three on windows they were never tuned on. The Default runs a profit factor of 1.24 on the year before the export and 0.97 on 2024 to mid-2025, with a worst quarter of −$1,084; Prop Safe 1.22 and 1.03, worst −$654 and seven of eleven quarters positive; Split Sessions 1.06 and 0.84, worst −$3,034. Split Sessions earns the most on the year and gives the most back outside it, because without the breakeven, trail and profit-side exits a loser runs to the full stop. The version page has the full holdout table and the equity curves.

The first build of this one meant for anyone other than us. It is an indicator, not a strategy — it takes no positions, so there is no P&L, no win rate and nothing to backtest. It lives under indicators.

What it does

Breaks each bar into its finest available sub-bars and books each sub-bar’s volume to the buy or the sell side depending on whether it ticked up or down — the tick rule — then plots the running difference. The body of each candle is that bar’s delta; the wicks are the intrabar high and low of the running total.

It approximates which side crossed the spread. Pine Script has no bid/ask-tagged trade data at any subscription tier, so no script on the platform can observe that directly, and this one infers it from price direction. It is not order flow, and the on-chart coverage row exists so the reading is never trusted blind.

Three fixes that changed what you see

Missing data no longer reads as zero. TradingView caps how far back intrabar history goes, and at 1-second resolution that window is short. The previous build booked every bar past the limit as a delta of zero and kept accumulating, so the cumulative line ran dead flat across the left of the chart — indistinguishable from a genuine absence of net flow. The plot now breaks there and the dashboard says MISSING. This was the main reason for the release.

Cumulative delta is drawn as candles. Plotting a per-bar histogram measured in tens against a cumulative line measured in thousands put two scales in one auto-scaled pane, which squashed the histogram to a hairline. Both readings are now the same series: candle body is the bar’s delta, candle position is the running total.

The session reset lands in the right place. The reset was documented as an exchange-session boundary and implemented as a calendar day. For CME futures those differ by six hours — the day opens at 18:00 ET the evening before. Exchange session is now the default, with calendar day still available.

Using it with a strategy

Seven series are exported and can be picked up by another script through its source input: bar delta, cumulative delta, delta %, buy volume, sell volume, coverage %, and a data-available flag.

Read the confirmed value, not the live one. Intrabar requests recalculate on every realtime tick, so the current bar’s delta moves until that bar closes. Reference the export one bar back, or gate on bar confirmation. Reading the live value will look excellent in a backtest and behave differently in real time.

What is not established

The price/delta divergence marks are unmeasured. They are the standard way cumulative delta is read, and they are why the script is worth publishing, but no hit rate has been established for them on any instrument and none is claimed. The build has been compiler-verified and reviewed; it has not yet been checked against TradingView’s own built-in volume delta on a live chart. It is marked draft until it has.

Renamed

Was Approx Volume Delta. “Aggressor” is the precise term for what the tick rule estimates — the side that crossed the spread — and it does not overstate anything the old name claimed.

A patch, and a narrow one: no strategy logic has changed. Not one line.

Tokyo Drift ships several configurations built on identical code, differing only in input defaults. v1.5.0 was published carrying the baseline set — the untuned development values — rather than the tuned Micro Gold set it was meant to ship with. This release moves 21 input defaults back to the intended values.

If you already added v1.5.0

TradingView does not reset your inputs when a script updates, so you are still running the old values until you act. Use Reset settings in the strategy’s settings dialog, or remove the script and re-add it.

The quickest check: Enable MACD Histogram Velocity Filter is on in the intended configuration and off in the baseline. Long TP is 3.5 rather than 2.5.

What moved

The MACD Histogram Velocity filter is on, and retuned to 18/22/7 with 12-bar velocity smoothing — the baseline had it switched off entirely, which is the single largest difference between the two.

Brackets: long target 3.5x ATR against a 2.35x stop, where the baseline ran 2.5x against 3.0x. That moves the long side from a below-1 reward:risk to about 1.5. Short stop 2.95x rather than 3.0x.

RSI oversold moves to 40 with an 18-bar extreme lookback. Asia opens at 7:00 PM ET rather than 6:30 PM; New York at 10:30 AM rather than 10:35 AM.

The remaining changes sit on filters that ship switched off in both configurations, so they alter nothing until those filters are enabled.

Measured

Micro Gold, 5-minute, one contract, $50,000 account, $1.20 per side and one tick of slippage, 24 May – 3 September 2026:

configuration	trades	net	win rate	PF	max drawdown
intended (this release)	235	+$21,318	71.49%	3.458	$550.40
baseline (what v1.5.0 shipped)	220	+$15,843	70.45%	2.715	$714.40

That window is also the window these defaults were tuned against, so it is a best case rather than an expectation. The longer-history figures and the concentration caveat on the version page are unchanged and still apply.

Patch release. One bug, no new settings, no default changed.

When the Reversal Engine armed, its budget was charged to the session whose trade had just been stopped out. That part was right. But the reversal trade itself drew its loss counter, its take-profit and stop, its breakeven and trail, and its end-of-day flatten from whichever session was live at the moment it filled.

Those are the same session almost always. They differ in exactly the case that matters: when the stopped-out trade ran past the end of its own window. An Asia position closing inside London’s hours armed with Asia’s budget and then filled as a London trade — so London absorbed the loss, supplied the bracket, and held the position until London’s end-of-day. With London set to one loss per session, a single Asia-armed reversal could exhaust London’s budget before London had taken a trade of its own.

A reversal now belongs to the session that armed it, for every purpose at once: budget, loss counter, bracket, breakeven, trail and flatten all agree.

Why it lasted this long

v3.3.0 fixed this same class of error for ordinary entries — trades were being credited to whichever session was live when they closed rather than when they opened. That fix predates the Reversal Engine, which arrived in v3.4.0, so the new code path was never covered by it.

It surfaced in the Tokyo Drift fork, which shares this structure, and was ported back here.

What changes on your chart

Nothing, unless you run the Reversal Engine with sessions whose windows overlap another session’s trading hours. If you do, expect slightly different trades — a session that was being locked out by a neighbour’s reversal can now take its own.

Verified against the existing parity fixture: 77 of 77 trades match exactly with the engine off, and no input default differs from v3.4.0.

Tokyo Drift’s first public release. It is a structural fork of the Drift VWAP Pullback line: where Drift gives each session a complete, independently tuned copy of every setting, Tokyo Drift applies one configuration to whichever sessions are enabled. Sessions still carry the clock, the direction toggles and the risk budget — everything else is shared.

Three independently tuned copies of one strategy is three chances to fit the past. One shared configuration has to earn its keep everywhere at once.

Five configurations

file	role
tokyo_drift_v1_5_0.pine	the published default — hand-tuned Micro Gold, verified live on TradingView
tokyo_drift_v1_5_0_default.pine	less-optimised alternate — untuned baseline plus the one validated gate
tokyo_drift_v1_5_0_prop.pine	robustness — profitable in all four periods tested
tokyo_drift_v1_5_0_prop_balanced.pine	funded-account evaluations (recommended of the two)
tokyo_drift_v1_5_0_prop_highwin.pine	highest hit rate, 79.1%

Identical code throughout; only input defaults differ, so moving between them is a settings change. The full parameter list for every one of them is on the version page, grouped the way the TradingView settings dialog groups them and copyable a section at a time.

The two prop files were selected on the trailing 365 days alone and carry no out-of-sample validation; prop.pine was chosen for surviving four independent periods. Both are covered in the prop-evaluation study.

The published default:

• Entry — RSI Reset inside a VWAP trend
• Direction gate — VWAP slope alone, with no requirement that price has already crossed
• Brackets — ATR multiples: long 2.5 target / 3.0 stop, short 3.0 / 3.0
• Filters on — RSI overbought/oversold, 15-point max candle range
• Trade management off — no breakeven, no trail
• Reversal Engine on — full stop-loss only, three-bar wait, confirming candle required

Read the results with the whole history in view

window	config	trades	net	win	PF	max DD
Full sample 2024-01 → 2026-08	Default	1,595	$13,452	52.0%	1.16	$7,011
Full sample	MGC-tuned	1,724	$8,582	47.9%	1.09	$13,669
Last 12 months	Default	679	$16,355	57.3%	1.35	$5,008
Last 12 months	MGC-tuned	716	$15,120	52.7%	1.30	$6,935
Last 6 months	Default	379	$18,817	65.2%	1.85	$1,092
Last 6 months	MGC-tuned	390	$20,519	62.1%	1.89	$1,645

One contract, $2.40 round turn, engine-measured on Databento MGC.

The full sample is thin, and the reason is concentration. On the default config the entire 2.5-year profit is two quarters:

	trades	$/trade	net
2026Q2 + 2026Q3	325	+$53.46	+$17,373
the other nine quarters	1,270	−$3.09	−$3,921

Nine of eleven quarters lose money, on both configurations. The obvious explanation — that the strategy needs a volatility floor and 2024’s gold was too quiet — was tested and does not hold: the split looks decisive until those two quarters are removed, after which it is statistically nothing, and the single highest-volatility quarter in the sample returned under $9 per trade. The edge is episodic and, so far, unexplained.

What the sample does not show is overfitting. Defaults frozen before 2025-26 existed lose 2024 just as hard as the tuned set, and the two track at pearson +0.954 across quarters while a different entry model on the same bars runs at −0.264. The quarter drives the result, not the settings.

The settings panel was rebuilt

Every distance now has one field and one unit, where several previously had a separate box per unit with all but one greyed out. 183 inputs down to 168.

That is not only tidying. A value parked in a hidden box is a value nobody checks, and in an earlier build exactly that turned a trailing distance into zero and quietly converted a profitable configuration into a losing one. One field cannot disagree with itself.

Also in this release

• Direction Gate — the price-side half of the trend test is now optional. It survived a walk-forward across eleven research stages and roughly 1,300 configurations, the only change that did. It is on in the published default; it defaults off in the tuned file only because each version has to reproduce its predecessor exactly at default settings before shipping, and that check is a release requirement rather than a recommendation.
• Reversal attribution fixed — a reversal’s budget was charged to the session that armed it while its loss went to whichever session it filled in, so one session could exhaust another’s budget before trading.
• Fractional contract sizes for the webhook payload, so CFD and spot venues that size in lots can be driven at all.

What this is not

One instrument over one sample. The ATR-scaled brackets will travel to another contract; the point-denominated values — candle-range filter, reversal bracket, trail distances — are gold-sized and will not. The published figures are engine-derived: the Python twin matches TradingView entry-for-entry at 97.6% on a paired export, but the tables above are not themselves TradingView runs. Backtest results are not a forecast, and with two profitable quarters out of eleven the next one is genuinely uncertain in both directions.

New York is now off by default. Asia and London stay on. NY is still profitable on a tuned configuration, but its losing trades have resisted every attempt to identify them in advance – every entry filter available, every tighter stop, and every holding-period cap was measured against a parity-validated engine over two separate windows, and all of them removed more in winners than they saved in avoided stops. Rather than ship a session whose weakest trades cannot be screened, NY is now a deliberate opt-in in the NY panel.

Reversal Engine, per session. When a trade is stopped out, the strategy can wait a set number of bars and take the opposite side with its own bracket. Which exit type arms it is configurable – a full stop-loss only, a breakeven/trail stop only, any stop, a winning exit, or any exit that isn’t the end-of-day flatten. Each session gets its own wait, its own take-profit and stop, and its own breakeven and trail, independent of the pullback leg.

The reversal budget is deliberately separate from Max Trades and Max Losses. A reversal is armed by a stop-out, and at the default of one loss per session that stop-out is exactly what exhausts the ordinary budget – so sharing it would have made the feature unreachable.

The trail engine has its own trigger. Previously a trail could only begin once breakeven had already fired. “Needs BE first” and “Trail Trigger (pts)” are now two independent gates, and either can be skipped. Leaving needs-BE on with a trigger of 0 reproduces the previous behaviour exactly, so existing setups are unchanged.

Per-session Enable Longs / Enable Shorts, and a per-session Max Extension Filter that blocks an entry already stretched more than a chosen number of standard deviations from that session’s own VWAP.

Bug fix – session guardrail attribution. Trade and loss counters were credited to whichever session was current when a trade closed, rather than the session that opened it. This was reachable on shipped settings: London’s end-of-day flatten lands on the exact minute NY’s window opens, so a London loss could consume NY’s loss budget before NY had placed a trade.

Bug fix – a pullback entry could take a pending reversal’s place. A normal signal arriving inside the reversal’s wait window won the race and the reversal never fired. A new per-session “Hold pullback entries during the wait” (on by default) prevents it. That is also the right default on its own merits: a pullback firing seconds after a stop-out is frequently a re-entry into the very move that just stopped the trade out.

Presentation. 201 inputs are now greyed out when they cannot affect anything, so an inert setting reads as inert. The trail input is relabelled Trail Distance – BE Offset is measured from your entry, Trail Distance from the best price reached, and the two words mean different anchors deliberately. Three new chart plots (breakeven trigger, breakeven lock, live trailing stop) and a unified session dashboard.

Backtest

MNQ, 5-minute, 2026-05-17 to 2026-08-30, 1 contract, $0.62 per contract commission, 1 tick slippage.

	default (Asia + London)	with NY opted in
trades	214	393
net profit	$10,228.64	$17,832.18
win rate	86.92%	83.46%
profit factor	3.777	3.071
max drawdown (closed trades)	$287.96	$405.22

One instrument over a single ~15-week window. Past performance does not guarantee future results.

Adds two things on top of v1.5.0’s per-day-of-week architecture:

• Stop & Reverse (SAR): on a per-day toggle, a tight initial stop-loss hit reverses the position into the opposite direction with its own fixed take-profit/stop-loss.
• 1-Hour EMA(9/21/50) stack gating: per-day toggle to switch trade direction based on whether the 1-hour EMA stack is bullish, bearish, or mixed at entry time.

Not yet production-trusted. The research behind these defaults came from a large parameter sweep with no out-of-sample validation – real overfitting risk, flagged directly rather than presented as a proven edge. Treat every SAR/gating number as a starting point for further testing, not a settled recommendation.

Bug fix: the “Current” and “Previous” VWAP Session Anchor options used to reset a session’s VWAP at that session’s trading-window start (when new entries are allowed) instead of that session’s real market-session start (when it actually begins trading). Retuning a session’s trading hours could silently move its VWAP anchor too. Fixed – Current and Previous now always resolve against each session’s real start, fully independent of the trading-window inputs.

Session Trading Window is now a clock-face range picker instead of a minutes-since-midnight number. EOD Close is now a plain hour/minute picker.

Added an optional VWAP standard-deviation band overlay showing the prevailing session trend at a glance, and an always-visible version badge so it’s clear which version’s settings are actually running.

The VWAP Session Anchor dropdown also gained “Asia” and “London” as fixed options – any session can now anchor to any other session’s real start by name.

Each day of the week (Sunday through Friday) trades independently: its own direction, entry/exit times, take-profit/stop-loss mode, breakeven, trailing stop, step-lock ratchet, and entry filters (gap size, prior-day trend, volatility, weekly trend).

Reviewed before shipping: the step-lock ratchet was fully working for long trades but missing entirely for short trades in the chart script, even though the backtest results it was validated against already included working short-side protection. Fixed before release.

Explores a different entry mechanism from Drift VWAP Pullback: instead of waiting for a specific pullback candle, it waits for RSI to swing to an extreme and then reset back through a confirming level – a more direct read of momentum exhaustion than candle color alone.

Reviewed and bug-fixed before release.

Promotes an independently-rebuilt pair of confluence filters (a trend-efficiency read and a momentum-velocity read) to replace an earlier filter whose original source couldn’t be confirmed as safe to ship. London runs the momentum-velocity filter, New York runs the trend-efficiency filter, Asia runs neither.

Ships with a broker-native webhook payload – stop-loss, take-profit, quantity, and strategy identity carried on every entry alert, no separate bracket alerts needed.

Confirmed trade-by-trade against a real TradingView export: 392 of 394 trades reproduced, entry and exit prices matching almost exactly.